Empirically Agent Based Modeling of Occupational Position Network in Japan

Monday, 11 July 2016: 09:30
Location: Hörsaal 27 (Main Building)
Oral Presentation
Hiroki TAKIKAWA, Tohoku University, Japan
Paolo PARIGI, Stanford University, USA
Social integration is one of the most classic themes in sociology and has become even more widely discussed in recent times. Despite growing academic interest in social integration, however, little is known about how social interactions are structured broadly in society due to the lack of adequate data and methodology to analyze broader network in a society. Previously, we developed the method generating macro-structural characteristics from egocentric network data, enabling to analyze the whole structure of social networks. Applying this method to a real data on egocentric networks in Japan that is generated by “position generator” focusing on ties with occupational positions, we obtained a occupational position network whose nodes represent occupational positions and whose edges represent similarity relationship between occupational positions.  Surprisingly, this empirical position network was largely divided by two major clusters.  The next step that we will take in this paper is to explicate generating mechanism of this segregated position network. For this purpose, we employ an empirically agent based modeling method. We incorporate two theoretical hypotheses into the model: status homophily and urbanity homophily. More concretely, in our model, two agents that are randomly matched with each other are more likely to be connected if statuses of them are similar, or if “urbanity” of their statuses are close. Specifically, the probability is determined by the following logistic functions,

exp(α+β1(-|prestigei-prestigej|)+β2(-|urbanityi-urbanityj|)/1+ exp(α+β1(-|prestigei-prestigej|)+β2(-|urbanityi-urbanityj|)

Both of status and “urbanity” parameters are not arbitrarily set up but can be measured empirically. The model well replicates the empirical position network in a certain range of parameter coefficients. From this result, it is concluded that the segregated position network in Japan would be generated from two major underlying processes: status homophily and urbanity homophily.